Forecasting maximum wind wave height from wave spectra using phase-resolving model and machine learning

The paper presents a combined methodology for the operational forecasting of maximum wave height, integrating the strengths of spectral wave models, phase-resolving simulations, and machine learning to address the core limitations inherent in each approach. The procedure begins with a frequency-directional wave spectrum obtained from the WAVEWATCH III model, which is subsequently transformed into a wavenumber field and used as initial conditions for the phase-resolving model TRIDWAVE. This step enables the generation of a realistic nonlinear wave field from which the target extreme parameter (maximum wave height) is extracted. To circumvent the prohibitive computational cost associated with repeatedly executing the phase-resolving model, a feedforward neural network was developed and trained to act as a fast surrogate, learning the mapping from input wave spectra to the corresponding maximum height values as calculated by TRIDWAVE. Validation experiments conducted for the Baltic Sea demonstrate that the trained network predicts maximum wave height with an average relative error of approximately 5 %. This result confirms the network’s capability to accurately infer key nonlinear statistics directly from linear spectral input.

1. Tolman HL. User manual and system documentation of WAVEWATCH III version 3.14. Technical Note. Camp Springs, MD: NOAA/NWS/NCEP/MMAB; 2009 May. 220 p.

2. Booij N, Ris RC, Holthuijsen LH. A third-generation wave model for coastal regions: 1. Model description and validation. Journal of Geophysical Research: Oceans.1999;104:7649–7666. https://doi.org/10.1029/98JC02622

3. The WAMDI Group. The WAM Model — A Third Generation Ocean Wave Prediction Model // Journal of Physical Oceanography. 1988. Vol. 18. P. 1775–1810. https://doi.org/10.1175/1520-0485(1988)018<1775:TWMTGO>2.0.CO;2

4. Chalikov DV, Bulgakov KY. Estimation of extreme waves statistics based on the statistics of significant wave height. Journal of Ocean Engineering and Marine Energy. 2017;3(4):417–423. https://doi.org/10.1007/s40722-017-0093-7

5. Chalikov DV, Bulgakov KY, Fokina KV. Interpretation of the spectral wave forecast model results using the phase-resolving model. Fundamental and Applied Hydrophysics. 2023;16(2):21–33. (In Russ.). https://doi.org/10.59887/2073-6673.2023.16(2)-2

6. Dias F, Brennan J, Ponce de Leon S, Clancy C, Dudley J. Local analysis of wave fields produced from hindcasted rogue wave sea states. In: Proceedings of the 34th International Conference on Ocean, Offshore and Arctic Engineering (OMAE2015); 2015 May 31-Jun 5; St. John’s, Newfoundland, Canada. New York: ASME; 2015. OMAE2015-41734. https://doi.org/10.1115/OMAE2015-41458

7. Chalikov DV, Fokina KV. Direct (phase-resolving) model of surface waves. Fundamental and Applied Hydrophysics. 2025;18(1):8–18. (In Russ.). https://doi.org/10.59887/2073-6673.2025.18(1)-1

8. Dommermuth DG, Yue DKP. A high-order spectral method for the study of nonlinear gravity waves. Journal of Fluid Mechanics. 1987;184:267–288. https://doi.org/10.1017/S002211208700288X

9. Adytia D, Saepudin D, Pudjaprasetya SR, Husrin S, Sopaheluwakan A. A deep learning approach for wave forecasting based on a spatially correlated wind feature, with a case study in the Java Sea, Indonesia. Fluids. 2022;7(1):39. https://doi.org/10.3390/fluids7010039

10. Castro E, Iuppa C, Musumeci RE, Foti E, Scandura P. Optimizing neural network training for nearshore sea state forecasts using maximum dissimilarity algorithm. Journal of Ocean Engineering and Marine Energy. 2025;11:1119–28. https://doi.org/10.1007/s40722-025-00426-5

11. Minuzzi FC, Farina L. Artificial neural networks ensemble methodology to predict significant wave height. Ocean Engineering. 2024;300:117479. https://doi.org/10.1016/j.oceaneng.2024.117479

12. Raissi M, Perdikaris P, Karniadakis GE. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics. 2019;378:686–707. https://doi.org/10.1016/j.jcp.2018.10.045

13. Zheng Z, Ma X, Huang X, Ma Y, Dong G. Wave forecasting within a port using WAVEWATCH III and artificial neural networks. Ocean Engineering. 2022;255:111475. https://doi.org/10.1016/j.oceaneng.2022.111475

14. Michalakes J, Dudhia J, Gill D, et al. The weather research and forecast model: Software architecture and performance. In: Zwiefhofer W, Mozdzynski G, eds. Use of High Performance Computing in Meteorology: Proceedings of the Eleventh ECMWF Workshop; 2004 Nov 1–5; Reading, United Kingdom. Singapore: World Scientific Publishing; 2005;156–168. https://doi.org/10.1142/9789812701831_0013

15. National Centers for Environmental Prediction/National Weather Service/NOAA/U.S. Department of Commerce, NCEP FNL Operational Model Global Tropospheric Analyses, continuing from July 1999. Research Data Archive at the National Center for Atmospheric Research, Computational and Information Systems Laboratory, Boulder, CO. 2000. https://doi.org/10.5065/D6M043C6

16. Fokina KV. Phase-resolving modeling of three-dimensional wind waves: dissertation … Candidate of Physical and Mathematical Sciences. St. Petersburg, 2024. 103 p.

17. NOAA Technical Memorandum: Amante, C. and B.W. Eakins, 2009. ETOPO1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC-24. National Geophysical Data Center, NOAA. http://doi.org/10.7289/V5C8276

18. Glorot X, Bordes A, Bengio Y. Deep sparse rectifier neural networks. Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics. 2011, PMLR15:315–323.

19. Luo G, Wang X, Zhao W, Tao S, Tang Z. ReLU Neural Networks and Their Training. Mathematics. 2026;14:39. http://doi.org/10.3390/math14010039

20. Rumelhart D., Hinton G., Williams R. Learning representations by back-propagating errors. Nature. 1986;323:533–536. http://doi.org/10.1038/323533a0

21. Kingma DP, Ba J. Adam: A method for stochastic optimization. Conference Paper at the 3rd International Conference for Learning Representations, San Diego, 2015. arXiv:1412.6980ss, https://doi.org/10.48550/arXiv.1412.6980